Automa'c design of digital synthe'c gene circuits Mario A. Marchisio and Joerg Stelling Department of Biosystems Science and Engineering ETH Zurich Anaheim, 15/06/10
Summary • Automa'c gene circuit design: the problem. • The Karnaugh map method in biology. • Comparison with a different design. • Circuit complexity and performance. • Conclusion and future work.
Automa'c gene circuit design: previous approaches. • Given the output, how to derive the corresponding circuit (structure and parameter values?) • Brute force op'miza'on via evolu'onary algorithm (François and Hakim, PNAS 101 , 580, 2004) • Similar implementa'ons: OptCircuit (Dasika and Maranas, BMC Syst. Bio. 2 , 24, 2008); Genetdes (Rodrigo et al ., Bioinforma6cs 23 , 1857, 2007). Problems: • Transcrip'on units as bio‐bricks (instead of parts). • Limited model (transla'on as single‐step event). • Double op'miza'on procedure: long computa'onal 'me. Looking for a different strategy.
Digital gene circuits • Input/Output rela'on fully described by a truth table. • The Karnaugh map method converts a truth table into a circuit scheme – no op'miza'on required. • Boolean gates due to promoter and RBS regula'on mechanisms. • Important applica'on as biosensors.
The Karnaugh map method Circuit structure in three layers – No op/miza/on required
Circuit characteris'cs • Ac'vated/Repressed Promoters and RBSs (Bintu et al. , Curr. Opin. Genet. Dev. 15 , 125, 2005; Isaacs et al., Nat. Biotech. 22 , 841, 2004) . • Pools of transcrip'on factors, sRNAs, and chemicals (M.A. Marchisio and J. Stelling, “Computa'onal design of synthe'c gene circuits with composable parts.” Bioinforma6cs , 24 , 1903, 2008). • A circuit takes up to four inputs (chemicals) and produces a single output (fluorescent protein).
Gate structure and new designs. • Riboswitches + sRNA on the RBS. • Promoters and RBS are controlled simultaneously.
Comparison with electronics • Every truth table corresponds to two Boolean formulas: CNF (POS) and DNF (SOP). • In electronics the minimal circuit is given by the formula with the lowest number of clauses of NOT opera'ons. • In biology several circuit schemes arise from the same Boolean formula. • How to define a minimal circuit in biology?
The complexity score • Regulatory factors maker more than gene number. • Only a handful of repressors and ac'vators is currently used. • Engineering new proteins is more difficult than synthesizing an'sense small RNAs. • Riboswitches simplify the structure of a gate. • We define as minimal the circuit with the lowest complexity score defined as S = 2 R−1 + 2 A−1 + n where: R, repressor number (>= 1); A, ac'vator number (>=1),n an'sense sRNA number • A circuit should avoid to re‐use the same kind of transcrip'on factors and prefer RBS controls to the promoter ones. • Riboswitches do not increase the circuit complexity.
Our tool • The truth table is the only input. • All the schemes compa'ble with POS and SOP formulas are computed (less than 1s up to 8s). • They are ranked according to their complexity score. • The user can choose a solu'on: this is built by parts, pools, and device composi'on and encoded in MDL (Model Defini'on Language) to be visualized in ProMoT (hkp://www.mpimagdeburg.mpg.de/projects/promot/).
Comparison with RNAi‐based design (Rinaudo et al. , Nat. Biotech. , 25 , 795, 2007) Our best solu'on Rinaudo’s solu'on (a+b)(A+c)(A+d) (acd)+(Ab) a b c d Circuit Score A R RNA Our best POS 2 0 2 0 Our tool found 15 designs with Our best SOP 4 2 1 1 complexity lower than 5. Rinaudo 5 0 0 5
How do these circuits work? • Circuit performance is es'mated through signal separa/on and transient calcula'on and depends both on structure and parameter values. &% ! & (% ! & min1 1234,35675*23-.80 '% ! & 0 1 Transient area Signal Settling time separation $% ! & max0 ! ! "!!! #!!!! #"!!! $!!!! )*+,-./0
A benchmark a b c d O 0 0 0 0 1 cd 00 01 11 10 • POS and SOP formulas have 0 0 0 1 0 ab 8 clauses each of which 0 0 1 0 0 00 1 0 1 0 contains 4 inputs. 0 1 0 0 0 01 0 1 0 1 • 48 possible schemes with S 1 0 0 0 0 11 1 0 1 0 varying from 20 to 2062. 0 0 1 1 1 10 0 1 0 1 0 1 0 1 1 1 0 0 1 1 0 1 1 0 1 a 1 0 1 0 1 b 1 1 0 0 1 0 1 1 1 0 c 1 0 1 1 0 1 1 0 1 0 Solu'on 1 1 1 1 0 0 d 1 1 1 1 1
Comparison of two possible solu'ons Solu'on 1 Solu'on 4 ! ) ! ) )%&"! )%&"! 1234-35675+23&.80 1234-35675+23&.80 ( min1 ( min1 ' ' # # max 0 max 0 ! ! ! " # $ ! " # $ *+,-&./0 *+,-&./0 ' ' %&"! %&"! Solu'on Rank Score A R RNA Gene Separa'on Transient 1 1 20 2 5 2 17 36 nM 3709 4 25 548 10 6 4 21 62.1 nM 5112 Higher complexity seems to guarantee beGer performance.
Improving the performance • The signal separa'on is mostly influenced by parameters belonging to the final gate . • Tuning only one parameter (the strength of the promoter in the final gate) the signal separa'on can be dras'cally amplified. ! * #%&"! 2345.46786,34&/91 "() " !() ! ! " # $ +,-.&/01 ' %&"! • Stochas'c algorithms can be avoided but a good set of default parameter values is required.
Conclusion and future work • Automa'c design of digital synthe'c gene circuits via the Karnaugh map method. • Circuit structure calcula'on does not require any op'miza'on procedure. • Theore'cal new design of Boolean gates where promoter and RBS are simultaneously. • Computer simula'ons show an unequivocal signal separa'on between 0/1 outputs with our choice of default parameter values. • Inser'on of other transla'on regula'on mechanisms. • Extension to eukaryo'c cells. • MAIN GOAL : Wet‐lab implementa'on of single gates and more complex circuits.
Recommend
More recommend